{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T08:50:02Z","timestamp":1772614202010,"version":"3.50.1"},"reference-count":32,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2024,5,9]],"date-time":"2024-05-09T00:00:00Z","timestamp":1715212800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Big Data"],"abstract":"<jats:p>In recent years, analyzing the explanation for the prediction of Graph Neural Networks (GNNs) has attracted increasing attention. Despite this progress, most existing methods do not adequately consider the inherent uncertainties stemming from the randomness of model parameters and graph data, which may lead to overconfidence and misguiding explanations. However, it is challenging for most of GNN explanation methods to quantify these uncertainties since they obtain the prediction explanation in a <jats:italic>post-hoc<\/jats:italic> and model-agnostic manner without considering the randomness of <jats:italic>graph data<\/jats:italic> and <jats:italic>model parameters<\/jats:italic>. To address the above problems, this paper proposes a novel uncertainty quantification framework for GNN explanations. For mitigating the randomness of graph data in the explanation, our framework accounts for two distinct data uncertainties, allowing for a direct assessment of the uncertainty in GNN explanations. For mitigating the randomness of learned model parameters, our method learns the parameter distribution directly from the data, obviating the need for assumptions about specific distributions. Moreover, the explanation uncertainty within model parameters is also quantified based on the learned parameter distributions. This holistic approach can integrate with any <jats:italic>post-hoc<\/jats:italic> GNN explanation methods. Empirical results from our study show that our proposed method sets a new standard for GNN explanation performance across diverse real-world graph benchmarks.<\/jats:p>","DOI":"10.3389\/fdata.2024.1392662","type":"journal-article","created":{"date-parts":[[2024,5,9]],"date-time":"2024-05-09T05:14:37Z","timestamp":1715231677000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":2,"title":["Quantifying uncertainty in graph neural network explanations"],"prefix":"10.3389","volume":"7","author":[{"given":"Junji","family":"Jiang","sequence":"first","affiliation":[]},{"given":"Chen","family":"Ling","sequence":"additional","affiliation":[]},{"given":"Hongyi","family":"Li","sequence":"additional","affiliation":[]},{"given":"Guangji","family":"Bai","sequence":"additional","affiliation":[]},{"given":"Xujiang","family":"Zhao","sequence":"additional","affiliation":[]},{"given":"Liang","family":"Zhao","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2024,5,9]]},"reference":[{"key":"B1","article-title":"Explainability techniques for graph convolutional networks","author":"Baldassarre","year":"2019","journal-title":"arXiv"},{"key":"B2","first-page":"610","article-title":"\u201cNetgan: generating graphs via random walks,\u201d","author":"Bojchevski","year":"2018"},{"key":"B3","first-page":"1050","article-title":"\u201cDropout as a bayesian approximation: representing model uncertainty in deep learning,\u201d","author":"Gal","year":"2016"},{"key":"B4","first-page":"131","article-title":"\u201cGnes: learning to explain graph neural networks,\u201d","author":"Gao","year":"2021"},{"key":"B5","first-page":"4094","article-title":"\u201cBayesian graph neural networks with adaptive connection sampling,\u201d","author":"Hasanzadeh","year":"2020"},{"key":"B6","doi-asserted-by":"publisher","first-page":"6968","DOI":"10.1109\/TKDE.2022.3187455","article-title":"Graphlime: local interpretable model explanations for graph neural networks","volume":"35","author":"Huang","year":"2022","journal-title":"IEEE Trans. Knowl. Data Eng"},{"key":"B7","article-title":"Adam: a method for stochastic optimization","author":"Kingma","year":"2014","journal-title":"arXiv"},{"key":"B8","article-title":"Semi-supervised classification with graph convolutional networks","author":"Kipf","year":"2016","journal-title":"arXiv"},{"key":"B9","doi-asserted-by":"crossref","first-page":"1010","DOI":"10.1145\/3534678.3539288","article-title":"\u201cSource localization of graph diffusion via variational autoencoders for graph inverse problems,\u201d","author":"Ling","year":"2022","journal-title":"Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining"},{"key":"B10","first-page":"21350","article-title":"\u201cDeep graph representation learning and optimization for influence maximization,\u201d","author":"Ling","year":""},{"key":"B11","doi-asserted-by":"crossref","first-page":"379","DOI":"10.1109\/ICDM51629.2021.00049","article-title":"\u201cDeep generation of heterogeneous networks,\u201d","author":"Ling","year":"2021","journal-title":"2021 IEEE International Conference on Data Mining (ICDM)"},{"key":"B12","doi-asserted-by":"publisher","first-page":"3099","DOI":"10.1007\/s10115-023-01863-0","article-title":"Motif-guided heterogeneous graph deep generation","volume":"65","author":"Ling","year":"","journal-title":"Knowl. Inf. Syst"},{"key":"B13","article-title":"Uncertainty decomposition and quantification for in-context learning of large language models","author":"Ling","year":"2024","journal-title":"arXiv"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1686","DOI":"10.1021\/ci300124c","article-title":"A bayesian approach to in silico blood-brain barrier penetration modeling","volume":"52","author":"Martins","year":"2012","journal-title":"J. Chem. Inf. Model"},{"key":"B15","doi-asserted-by":"publisher","first-page":"105043","DOI":"10.1016\/j.isci.2022.105043","article-title":"Edgeshaper: bond-centric shapley value-based explanation method for graph neural networks","volume":"25","author":"Mastropietro","year":"2022","journal-title":"Iscience"},{"key":"B16","doi-asserted-by":"publisher","first-page":"80","DOI":"10.3389\/fenvs.2015.00080","article-title":"Deeptox: toxicity prediction using deep learning","volume":"3","author":"Mayr","year":"2016","journal-title":"Front. Environ. Sci"},{"key":"B17","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41598-021-84854-x","article-title":"Dropconnect is effective in modeling uncertainty of bayesian deep networks","volume":"11","author":"Mobiny","year":"2021","journal-title":"Sci. Rep"},{"key":"B18","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.neucom.2022.11.049","article-title":"A general framework for quantifying aleatoric and epistemic uncertainty in graph neural networks","volume":"521","author":"Munikoti","year":"2023","journal-title":"Neurocomputing"},{"key":"B19","article-title":"Bayesian graph convolutional neural networks using non-parametric graph learning","author":"Pal","year":"2019","journal-title":"arXiv"},{"key":"B20","first-page":"10772","article-title":"\u201cExplainability methods for graph convolutional neural networks,\u201d","author":"Pope","year":"2019","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"key":"B21","first-page":"5281","article-title":"\u201cDirect uncertainty prediction for medical second opinions,\u201d","author":"Raghu","year":"2019"},{"key":"B22","article-title":"Interpreting graph neural networks for nlp with differentiable edge masking","author":"Schlichtkrull","year":"2020","journal-title":"arXiv"},{"key":"B23","doi-asserted-by":"publisher","first-page":"7581","DOI":"10.1109\/TPAMI.2021.3115452","article-title":"Higher-order explanations of graph neural networks via relevant walks","volume":"44","author":"Schnake","year":"2021","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1936","DOI":"10.1021\/acs.jcim.6b00290","article-title":"Computational modeling of \u03b2-secretase 1 (bace-1) inhibitors using ligand based approaches","volume":"56","author":"Subramanian","year":"2016","journal-title":"J. Chem. Inf. Model"},{"key":"B25","article-title":"Graph attention networks","author":"Veli\u010dkovi\u0107","year":"2017","journal-title":"arXiv"},{"key":"B26","first-page":"12225","article-title":"Pgm-explainer: probabilistic graphical model explanations for graph neural networks","volume":"33","author":"Vu","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B27","article-title":"Batchensemble: an alternative approach to efficient ensemble and lifelong learning","author":"Wen","year":"2020","journal-title":"arXiv"},{"key":"B28","author":"Xu","year":"2018"},{"key":"B29","first-page":"9240","article-title":"Gnnexplainer: generating explanations for graph neural networks","volume":"32","author":"Ying","year":"2019","journal-title":"Adv. Neural Inf. Process. Syst"},{"key":"B30","doi-asserted-by":"crossref","first-page":"2104","DOI":"10.1145\/3442381.3449818","article-title":"\u201cTg-gan: continuous-time temporal graph deep generative models with time-validity constraints,\u201d","author":"Zhang","year":"2021","journal-title":"Proceedings of the Web Conference 2021"},{"key":"B31","doi-asserted-by":"crossref","first-page":"5829","DOI":"10.1609\/aaai.v33i01.33015829","article-title":"\u201cBayesian graph convolutional neural networks for semi-supervised classification,\u201d","author":"Zhang","year":"2019","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 33"},{"key":"B32","first-page":"12827","article-title":"Uncertainty aware semi-supervised learning on graph data","volume":"33","author":"Zhao","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst"}],"container-title":["Frontiers in Big Data"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2024.1392662\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,5,9]],"date-time":"2024-05-09T05:14:45Z","timestamp":1715231685000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2024.1392662\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,5,9]]},"references-count":32,"alternative-id":["10.3389\/fdata.2024.1392662"],"URL":"https:\/\/doi.org\/10.3389\/fdata.2024.1392662","relation":{},"ISSN":["2624-909X"],"issn-type":[{"value":"2624-909X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,5,9]]},"article-number":"1392662"}}